BareML™: A Toolkit for Embedded Engineers

---

Reservoir Aided Inertial Navigation

• Embedded ML for high-speed fusion and filtering of navigation data, including raw IMU signals and alternate PNT signals of opportunity
• Algorithms designed to minimize latency and power consumption while operating on standard microcontrollers associated with IMU boards
• Ongoing development funded in part by DoD

Reservoir-Augmented Control and Health system

• Solution for real-time ML processing of complex sensor signals
• Developed to interpret SansEC sensor information (created by NASA, SansEC sensors are wireless, passive, and contactless, making them ideal for aerospace applications)
• ReACH performs data fusion, inference, state estimation, and classification tasks on embedded low-power compute hardware
• Ongoing development supported by NASA; initial development funded in part by DoD

PX4 branch implementing ML-based adaptive control and health monitoring

• Replaces PX4 code modules with ML models that are able to retrain in flight as needed
• Runs on standard autopilot hardware via a software push--no companion computer required
• Exports ML-derived health and status information to custom QGC display plugin
• Ongoing development funded in part by Ohio Third Frontier and DoD

Reservoir-Augmented Power, Propulsion, and Thermal management system

• ReACH software applied to modeling and control of hybrid powertrains
• Creates and evaluates Digital Twins of PPT subsystems and overall system performance using an onboard microcontroller
• Exports ML-derived health and status information to an intuitive operator display
• Ongoing development funded in part by DoD

Customized advanced edge algorithms for

• Signal characterization and processing
• High-speed data fusion
• Health and status monitoring and predictive health analysis
• ML-augmented adaptive control